Summary:
- The article discusses the methodological challenges of evaluating Artificial Intelligence (AI) within educational contexts, specifically critiquing the current reliance on "evidence-based" frameworks.
- It argues for a shift away from standardized, narrow performance metrics toward more nuanced, experimental approaches that account for the sociotechnical complexity of classroom environments.
- The author emphasizes that AI evaluation in education should be viewed as a situated, iterative process rather than a static validation of technological efficacy.